oComm: Overlapping Community Detection in Multi-View Brain Network

Higher Order Connection Enhanced Community Detection in Adversarial Multiview Networks

Community detection in multiview networks has drawn an increasing amount of attention in recent years. Many approaches have been developed from different perspectives. Despite the success, the problem of community detection in adversarial multiview networks remains largely unsolved. An adversarial multiview network is a multiview network that suffers an adversarial attack on community detection in which the attackers may deliberately remove some critical edges so as to hide the underlying community structure, leading to the performance degeneration of the existing approaches. To address this problem, we propose a novel approach, called higher order connection enhanced multiview modularity (HCEMM). The main idea lies in enhancing the intracommunity connection of each view by means of utilizing the higher order connection structure. The first step is to discover the view-specific higher order Microcommunities (VHM-communities) from the higher order connection structure. Then, for each view of the original multiview network, additional edges are added to make the nodes in each of its VHM-communities fully connected like a clique, by which the intracommunity connection of the multiview network can be enhanced. Therefore, the proposed approach is able to discover the underlying community structure in a multiview network while recovering the missing edges. Extensive experiments conducted on 16 real-world datasets confirm the effectiveness of the proposed approach.

Investigating the spectral features of the brain meso-scale structure at rest

Recent studies provide novel insights into the meso-scale organization of the brain, highlighting the co-occurrence of different structures: classic assortative (modular), disassortative, and core-periphery. However, the spectral properties of the brain meso-scale remain mostly unexplored. To fill this knowledge gap, we investigated how the meso-scale structure is organized across the frequency domain. We analyzed the resting state activity of healthy participants with source-localized high-density electroencephalography signals. Then, we inferred the community structure using weighted stochastic block-model (WSBM) to capture the landscape of meso-scale structures across the frequency domain. We found that different meso-scale modalities co-exist and are diversely organized over the frequency spectrum. Specifically, we found a core-periphery structure dominance, but we also highlighted a selective increase of disassortativity in the low frequency bands (<8 Hz), and of assortativity in the high frequency band (30-50 Hz). We further described other features of the meso-scale organization by identifying those brain regions which, at the same time, (a) exhibited the highest degree of assortativity, disassortativity, and core-peripheriness (i.e., participation) and (b) were consistently assigned to the same community, irrespective from the granularity imposed by WSBM (i.e., granularity-invariance). In conclusion, we observed that the brain spontaneous activity shows frequency-specific meso-scale organization, which may support spatially distributed and local information processing.

Adaptively Weighted Multiview Proximity Learning for Clustering

Recently, the proximity-based methods have achieved great success for multiview clustering. Nevertheless, most existing proximity-based methods take the predefined proximity matrices as input and their performance relies heavily on the quality of the predefined proximity matrices. A few multiview proximity learning (MVPL) methods have been proposed to tackle this problem but there are still some limitations, such as only emphasizing the intraview relation but overlooking the inter-view correlation, or not taking the weight differences of different views into account when considering the inter-view correlation. These limitations affect the quality of the learned proximity matrices and therefore influence the clustering performance. With the aim of breaking through these limitations simultaneously, a novel proximity learning method, called adaptively weighted MVPL (AWMVPL), is proposed. In the proposed method, both the intraview relation and the inter-view correlation are considered. Besides, when considering the inter-view correlation, the weights of different views are learned in a self-weighted scheme. Furthermore, through an adaptively weighted scheme, the information of the learned view-specific proximity matrices is integrated into a view-common cluster indicator matrix which outputs the final clustering result. Extensive experiments are conducted on several synthetic and real-world datasets to demonstrate the effectiveness and superiority of our method compared with the existing methods.